Recognition of Control Chart Patterns Using Self Organization Models

Abstract

The study proposes the use of self organizing neural networks for accurate and speedy detection of control chart patterns in order to achieve tight control of the process and ensuring good product quality. Control charts used in statistical process control can exhibit six principal types of patterns. Apart from the normal patterns, all other patterns indicate abnormalities in the process, which must be corrected to bring the process under control. Hence, accurate identification of control chart pattern is essential in modern industry. Unlike conventional tools that require a prior knowledge of the problem, the network classifies in a pure intuitive manner. Further the network is noise-tolerant. The training and classification results of the network are presented.